The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): This skill ingests untrusted data from JSON and Schema files and possesses a file-write capability, which could be exploited to overwrite sensitive files if the agent is manipulated.
Ingestion points: scripts/json_validator.py reads user-provided file paths in validate_json and validate_schema.
Boundary markers: None. Content is parsed directly into Python objects without delimiters or instructions to ignore embedded commands.
Capability inventory: The script includes a file-writing operation via the --output argument (json_validator.py line 73).
Sanitization: None. Data is processed using the standard json library, which does not filter content for instructions that may influence the agent's behavior.
[Unverifiable Dependencies & Remote Code Execution] (LOW): The script relies on the external jsonschema library. While it is a standard library for this task, it is an external dependency not included in the defined list of trusted sources.

json-validator